Storage layer

Chive uses multiple specialized databases, each optimized for different access patterns:

Database	Purpose	Documentation
PostgreSQL	Primary data store for eprints, reviews, profiles	PostgreSQL
Elasticsearch	Full-text search, faceted navigation	Elasticsearch
Neo4j	Knowledge graph, citations, field hierarchy	Neo4j
Redis	Caching, rate limiting, sessions	Redis

Architecture

All storage adapters implement consistent interfaces for dependency injection:

interface IStorageBackend {
  connect(): Promise<void>;
  disconnect(): Promise<void>;
  healthCheck(): Promise<boolean>;
}

Data flow

1. Write path: Firehose events → Services → PostgreSQL → Async sync to ES/Neo4j
2. Read path: API → Services → PostgreSQL/ES/Neo4j (based on query type)
3. Cache: Redis sits in front of expensive queries

See also

• Core services - How services use storage adapters
• Observability - Database monitoring and metrics